A glove-controlled tracked robot responds to a raised middle finger by raising its own mechanical finger. The maker says the build uses two Arduino Nano boards, a motion sensor in the glove, four drive motors and a servo for the finger.
The project was shared in a maker's Reddit post.
A remote-controlled robot with a very specific talent: outsourcing your rude gestures.
This Fish Does Not Exist is a framed e-paper display that creates textbook-style pages about imaginary fish. Software generates each fish's appearance, while an on-device AI model invents its scientific description. Every few minutes a new specimen appears, drawn stroke by stroke, with its description revealed gradually.
Maker Jim Bruges built it around an Arduino UNO Q and a 7.5-inch e-paper panel. Gemma 3 1B runs locally, so after setup the device can keep producing its fictional field guide without cloud services.
Instead of waiting for scientists to discover a new species, the frame simply invents one. The field guide never runs out of fish.
The Brass Thumb Pomodoro is a steampunk-style desk timer that rewards a completed 25-minute work session by popping a thumbs-up figure out of its case. A side switch selects work or rest mode, with five-minute breaks between sessions.
Maker Krzysztof Jamroz built the USB-powered device with Arduino electronics and 3D-printed parts. Assembly instructions and code are available on GitHub.
Finish 25 minutes of work and a mechanical thumb pops up to congratulate you. At least one thing on your desk is guaranteed to appreciate your effort.
P.A.R. is a 37×18 display made from 666 two-sided square tiles. Instead of giving every pixel its own actuator, a single robot moves behind the panel and physically flips individual tiles as needed to draw monochrome pixel art.
Artwork can be submitted through a website, and the machine reproduces it on the physical grid. The CAD models and code are open source. The project was created by Zimm.
Every time the display updates, a robot has to travel around behind the screen and physically flip the pixels one by one. Refreshing the image has become an actual mechanical chore.
A 1970s Texas Instruments Silent 700, which combines a keyboard and thermal printer, has been connected to a local AI so that answers arrive on paper. The device, called Termi3, prints a response two or three seconds after a question is typed. It was built by Cameron Coward.
Earlier versions needed an internet connection to query Wolfram Alpha, but Termi3 runs a 2 GB local LLM on the Arduino and works entirely offline. Coward notes that its answers are sometimes reasonable and accurate, and other times baffling or hilarious.
A modern local AI does the thinking, but the output device is still a 1970s paper terminal. The intelligence got newer while every reply still arrives as a printout.
Edge AI Balcony Guard detects visiting pigeons locally and sweeps a water gun to drive them away. It is designed to deter pigeons without a camera and without harming them, helping prevent droppings on balconies.
Creator Hiroshi Nakata built it with Arduino, a servo, and Solist-AI, and published it on ProtoPedia on September 23, 2026.
The final form of balcony security is an AI-controlled automatic water gun. From a pigeon's perspective, the equipment is taking this more seriously than the resident.
A homemade alarm clock gives the sleeper five seconds to stop the alarm before a loaded banknote is fed into a shredder. Instead of merely continuing to make noise, the device makes oversleeping carry a physical monetary consequence.
The Arduino-based build was published by the Japanese maker channel Mechatro on September 12, 2026.
The punishment for snoozing is not a louder alarm but the physical loss of a banknote five seconds later. Waking up now has a financial deadline.
Newson's Electronics' Self Catching Projectile Machine launches a ball bearing, moves a catcher to the predicted landing point, and catches its own projectile.
It reads the launch angle and calculates the range using projectile speed and gravity. An Arduino then moves the catcher along a rail to the calculated target position and returns the mechanism to its home position for the next shot.
Put the thrower and catcher into the same machine and even a one-person game of catch no longer needs the person.
James' Programming built a physical illusion inspired by Portal: throw an object into one frame and a matching object appears to shoot out of the other. There is no video compositing; the entrance detects the incoming object while a flywheel launcher at the exit fires a duplicate.
Time-of-flight sensors detect the approaching object early enough to compensate for the launcher's mechanical delay. The build took about nine months, and the full detection and automatic realignment sequence was still finicky, succeeding roughly one time in twenty.
Instead of inventing teleportation, it catches the object at one end and aggressively launches a duplicate from the other. Portal, but fully compliant with physics.
After building a fully functional 7:1-scale Arduino UNO, UncleStem made a matching 7:1-scale version of the classic beginner turtle robot. The oversized parts mimic an UNO, L298N driver, ultrasonic sensor and jumper wires, while the internals are replaced with appropriately beefy hardware such as 12–24 V ride-on-toy motors and a pair of 300 W motor drivers.
The giant robot still behaves like the original classroom build: it pans an ultrasonic sensor, detects obstacles and drives around them.
The obvious question after building a giant Arduino is “what is it for?” The answer was not to shrink the Arduino—it was to enlarge the entire beginner robot around it.
A small robot car has been built to detect a face with its camera, turn until that face is centered and then drive toward the person it sees. The project upgrades an Elegoo robot car kit with an Arduino UNO Q and webcam, using computer vision to locate faces and adjust steering. It was created by Iulia Feroli and featured on Arduino's official blog.
Face detection is familiar on a screen; connecting the detection result directly to motion gives the same software a much stronger physical presence.
Face detection feels harmless on a screen, then you add wheels and it starts closing the distance. Giving AI mobility really changes the mood.
A remote-controlled robot with a very specific talent: outsourcing your rude gestures.